Re: [PATCH net 1/2] ipv4: fix data races in fib_alias_hw_flags_set

From: Ido Schimmel <idosch@idosch.org>
To: Eric Dumazet <eric.dumazet@gmail.com>
Cc: "David S . Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>, netdev <netdev@vger.kernel.org>,
	Eric Dumazet <edumazet@google.com>,
	Ido Schimmel <idosch@nvidia.com>,
	syzbot <syzkaller@googlegroups.com>
Subject: Re: [PATCH net 1/2] ipv4: fix data races in fib_alias_hw_flags_set
Date: Wed, 16 Feb 2022 23:30:59 +0200	[thread overview]
Message-ID: <Yg1tExg4IGAKm7Bq@shredder> (raw)
In-Reply-To: <20220216173217.3792411-1-eric.dumazet@gmail.com>

On Wed, Feb 16, 2022 at 09:32:16AM -0800, Eric Dumazet wrote:
> From: Eric Dumazet <edumazet@google.com>
> 
> fib_alias_hw_flags_set() can be used by concurrent threads,
> and is only RCU protected.
> 
> We need to annotate accesses to following fields of struct fib_alias:
> 
>     offload, trap, offload_failed
> 
> Because of READ_ONCE()WRITE_ONCE() limitations, make these
> field u8.

[...]

> Fixes: 90b93f1b31f8 ("ipv4: Add "offload" and "trap" indications to routes")
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Cc: Ido Schimmel <idosch@nvidia.com>
> Reported-by: syzbot <syzkaller@googlegroups.com>

Reviewed-by: Ido Schimmel <idosch@nvidia.com>

Thanks, this is better than taking RTNL

> ---
>  net/ipv4/fib_lookup.h    |  7 +++----
>  net/ipv4/fib_semantics.c |  6 +++---
>  net/ipv4/fib_trie.c      | 22 +++++++++++++---------
>  net/ipv4/route.c         |  4 ++--
>  4 files changed, 21 insertions(+), 18 deletions(-)
> 
> diff --git a/net/ipv4/fib_lookup.h b/net/ipv4/fib_lookup.h
> index e184bcb1994343c00914e09ab728ae16c4d23dc8..78e40ea42e58d930b3439d497de2b9e15fe45706 100644
> --- a/net/ipv4/fib_lookup.h
> +++ b/net/ipv4/fib_lookup.h
> @@ -16,10 +16,9 @@ struct fib_alias {
>  	u8			fa_slen;
>  	u32			tb_id;
>  	s16			fa_default;
> -	u8			offload:1,
> -				trap:1,
> -				offload_failed:1,
> -				unused:5;
> +	u8			offload;
> +	u8			trap;
> +	u8			offload_failed;

There is a 5 bytes hole here, so this shouldn't increase the struct and
it should still fit in one cache line

>  	struct rcu_head		rcu;
>  };